The present invention relates generally to data sensing within semiconductor memory devices and, more particularly, to method and apparatus for initializing reference cells of a toggle switched memory device.
Magnetic (or magneto-resistive) random access memory (MRAM) is a non-volatile random access memory technology that could replace the dynamic random access memory (DRAM) and/or static random access memory (SRAM) as the standard memory for computing devices. The use of MRAM as a non-volatile RAM would allow for “instant on” systems that come to life as soon as the system is turned on, thus saving the amount of time needed for a conventional PC, for example, to transfer boot data from a hard disk drive to volatile DRAM during system power up.
A magnetic memory element (also referred to as a tunneling magneto-resistive, or TMR device) includes a structure having ferromagnetic layers separated by a non-magnetic layer (barrier), and arranged into a magnetic tunnel junction (MTJ). Digital information is stored and represented in the memory element as directions of magnetization vectors in the magnetic layers. More specifically, the magnetic moment of one magnetic layer (also referred to as a reference layer) is maintained in a preassigned direction, while the magnetic moment of the magnetic layer on the other side of the tunnel barrier (also referred to as a “free” layer) may be switched during operation between the same direction and the opposite direction with respect to the fixed magnetization direction of the reference layer. The orientations of the magnetic moment of the free layer adjacent to the tunnel junction are also known as “parallel” and “antiparallel” states, wherein a parallel state refers to the same magnetic alignment of the free and reference layers, while an antiparallel state refers to opposing magnetic alignments therebetween.
Depending upon the magnetic state of the free layer (parallel or antiparallel), the magnetic memory element exhibits two different resistance values in response to a voltage applied across the tunnel junction barrier. The particular resistance of the TMR device thus reflects the magnetization state of the free layer, wherein resistance is typically “low” when the magnetization is parallel, and “high” when the magnetization is antiparallel. Accordingly, a detection of changes in resistance allows a MRAM device to provide information stored in the magnetic memory element (i.e., a read operation). There are different methods for writing a MRAM cell; for example, a Stoner-Wohlfarth astroid MRAM cell is written to through the application of fields to exceed a critical curve or stability threshold, in order to magnetically align the free layer in a parallel or antiparallel state. The free layer is fabricated to have a preferred axis for the direction of magnetization called the “easy axis” (EA), and is typically set by a combination of intrinsic anisotropy, strain induced anisotropy, and shape anisotropy of the MTJ.
When a sufficiently large current is passed through both a wordline and a bitline of the MRAM, the combined fields of these currents at the intersection of the write and bit lines will switch the magnetization of the free layer of the particular MTJ located at the intersection of the energized write and bit lines. The current levels are selected such that the combined fields exceed the switching threshold of the free layer. For a Stoner-Wohlfarth astroid MRAM structure, the EA is aligned with the orientation of either the bitline or the wordline.
As the lateral dimension of an MRAM device decreases, several problems can occur. First, the switching field increases for a given shape and film thickness, thus requiring a larger magnetic field for switching. Second, the total switching volume is reduced such that the energy barrier for reversal also decreases, wherein the energy barrier refers to the amount of energy needed to switch the magnetic moment vector from one state to the other. The energy barrier determines the data retention and error rate of the MRAM device, and thus unintended reversals can occur due to thermal fluctuations if the barrier is too small. Furthermore, with a small energy barrier it becomes extremely difficult to selectively switch a single MRAM device in an array without inadvertently switching other MRAM devices. Thirdly, the shape anisotropy becomes dominant and the switching field becomes more sensitive to shape variations as the MRAM devices decrease in size.
In this regard, there has also been introduced an MRAM device in which the free layer of ferromagnetic material includes multiple (e.g., two) ferromagnetic layers separated by a nonmagnetic coupling layer. Due to magnetostatic coupling, the magnetic moments of the two ferromagnetic layers are antiparallel to one another. This configuration allows for a different method of writing that improves selectivity. An exemplary configuration aligns the axis of intrinsic magnetic anisotropy at a 45° angle with respect to the orientation of the word and bitlines. The device can be patterned to include shape anisotropy, and in an exemplary configuration is also aligned at a 45° angle with respect to the orientation of the word and bitlines.
More specifically, the writing method relies on a toggle or “spin-flop” phenomenon that gently rotates the magnetic moment vectors of the two ferromagnetic layers so they exchange directions. In the X-Y field plane, the fields follow a closed trajectory or “toggle-box” that encloses a critical point called the spin-flop point. The magnitudes of the required fields are dependent on the location of this spin-flop point. Current waveforms applied to the wordline and bitline in a timed sequence induce a magnetic field trajectory which reliably toggles the state of the multifilm free layer, such that the magnetization of the film closest to the tunnel barrier will switch direction (i.e., “toggle”), and at remanence the partner film in the free layer maintains an antiparallel magnetization to the aforementioned film. Additional information regarding toggle based MRAM writing may be found in, for example, U.S. Pat. No. 6,545,906 to Savtchenko, et al. Thus, in order to write a toggle based MTJ into a given state, it is first necessary to determine its initial state and then toggle it if required.
A common method for reading conventional MTJs is to apply a voltage and compare the resulting current to that of one or more reference MTJs of known state. Accordingly, such reference cells must first be initialized to a known state. However, this is difficult in a toggle switched memory since no such independent reference exists for determining the initial state of the reference cells themselves in the first place. That is, the write operation of a toggle mode device operates by changing the logical state of the cell. In other words, if the state of a reference cell is not known in the first place, toggling the cell simply changes its state without knowledge of what the previous or current state of the cell is. Accordingly, it would be desirable to be able to first determine an initial state of the reference cells in a toggle switched memory, so that the references may thereafter be initialized to desired values for operation of the memory device.